"Robot + AI" accelerates research on solar cell materials

Solar energy is one of the important energy sources of the future, but making more efficient solar cells requires finding new and better materials. Recently, researchers from Osaka University proposed a solution in a study published in JACS Au that can automate key experimental and analytical processes, thereby greatly speeding up the research of solar materials.

Traditional solar cells are made of inorganic semiconductors such as silicon and gallium, but the next generation of solar cells needs to make breakthroughs in cost, weight and safety. In addition, existing solar cells often contain toxic lead, so there is a need to find less toxic alternative materials. However, the current process of researching new materials is done manually, which is expensive and time-consuming.

To address this problem, the researchers developed a unique robotic measurement system capable of performing optical absorption spectroscopy, optical microscopy, and time-resolved microwave conductivity analysis. The key feature of this system is that it can be automated to efficiently perform multiple experimental and analytical processes. With the help of this automated system, the researchers evaluated a total of 576 different thin-film semiconductor samples.

Lead author Chisato Nishikawa noted: "Current solar cells are mainly made of inorganic semiconductors such as silicon and gallium, but the next generation of solar cells needs to have breakthroughs in cost, weight and toxicity. Although perovskite solar cells are efficient enough to compete with silicon solar cells, they contain toxic lead."

In this study, the researchers studied solution-processed lead-free solar cells composed of four elements, Cs, Bi, Sb and I, with a wide range of composition and process parameters. In order to gain a deeper understanding of the properties of these materials and automate the entire experimental process, the researchers used AI-related technologies, especially machine learning technologies, to analyze the data generated by the experiments.

"In recent years, machine learning has been extremely helpful in better understanding the properties of materials," said senior author Akinori Saeki. "These studies require a large amount of experimental data, and combining automated experiments with machine learning techniques is an ideal solution."

In the future, the researchers hope to automate more of the experimental process, making it easier to explore completely new materials. As Chisato Nishikawa points out: "This method is very suitable for exploring areas where there is no existing data."

So far, the research team's robotic system has achieved the results they expected. The measurement process is fully automated and highly accurate, and can complete the work in one-sixth of the time usually required. This automated system makes the task of finding efficient and non-toxic solar materials easier, providing more hope for the future of solar energy. The synergy of robots and artificial intelligence may bring solar energy one step closer to us.